Electromagnetic Information Theory Meets Artificial Intelligence: Fundamentals, Recent Advances, and Future Opportunities

Electromagnetic information theory (EIT) serves as a critical bridge between the foundational principles of information theory and the physical characteristics of electromagnetic (EM) wave propagation, offering a unified framework for understanding and optimizing wireless communications, sensing, and computational EM systems. Recent advances in artificial intelligence (AI) have unveiled a large number of new opportunities for EIT-based architectures, from physically-aware system analysis to adaptive design strategies. The integration of EIT and AI enables the development of more precise, flexible, and intelligent EM systems, overcoming long-standing bottlenecks and paving the way for next-generation wireless technologies. This review comprehensively discusses the latest advancements and future trends at the convergence of EIT and AI. Firstly, the fundamental principles of EIT framework are systematically introduced, including novel EIT-based multiple-input multiple-output (MIMO) paradigms, EM channel modeling and analysis, as well as some emerging techniques such as reconfigurable intelligent surfaces (RISs), wavefront shaping, and near-field communications. Subsequently, the latest developments at the intersection of EIT and AI are also presented, demonstrating how advanced AI-driven approaches are revolutionizing system performance, adaptability, and robustness across a range of EIT applications. Finally, future trends and promising prospects for the implementation of AI-based EIT methods are outlined, aiming to inspire further interdisciplinary research and innovation in this rapidly evolving field.